Comparison of the Signal-to-Noise Sensitivity of Different False Discovery Rate Methods
Murugavel M, Sullivan J, Helmer K
Worcester Polytechnic Institute
The FDR (False Discovery Rate) has become a popular method of correcting for multiple comparisons in fMRI data analysis. These methods differ in their assumptions of the underlying correlation structure between fMRI voxels. Significant difference were obtained when comparing the result of five different FDR algorithms on the p-value results of a hand-flex task fMRI study. The algorithms were also applied to a data set that simulated different signal-to-noise levels. It was found that the differences in number of voxels deemed to be active was correlated tothe degree of assumptions made about the data correlation structure.